abstract : we study a class of (possibly infinite ...infinite-dimensional lie algebras, tr1ere e;
TRANSCRIPT
CLASSIFICATION AND CONSTRUCTION
OF
OUASISIMPLE LIE ALGEBRAS
Raphael H0EGH-KROHN(*) Bruno TORRESANI(,....l
Abstract : We study a class of (possibly infinite-dimensional) Lie algebras,
called the Quasisimple Lie algebras (QSLA's), and generalizing semisimple and affine Kac- Moody Lie algebras. They are characterized by the existence of a finite-dimensional Cartan subalgebra, a non-degenerate symmetric ad-invariant Killing form, and nilpotent rootspaces attached to non-isotropic roots. We are then able to derive a classification theorem for the possible quasisimple root systems; moreover, we construct explicit realizations of some of them as current algebras, generalizing the affine loop algebras.
(*) Universitetet i Oslo, Matematisk lnstltutt. Blmdern Oslo 3, Oslo, Norway
(**) Allocataire du M.R.T.
( +) Laboratoire pro pre du CNRS - LP 7061
I. INTRODUCTION:
Infinite-dimensional Lie algebras and their repr-esentations have been
st-IO\·Vn to be a '·lery e;.::citing and povverful tool for U1e investigation of many
apparently disconnected fields in r·nau·1ernatics and rnatr1ematica1 physics, like
combinatorial identities [2], [3].. non-linear partial differential equations of
"soliton type" [4].. dual resonance models and string theory [51 and anomalies
in quanturn gauge theories [6] for instance.
In fact_. it is hoped that tr1ey allow a better understanding of tr1e
remarkable interrelations bet'·,·veen U1e~;e different fields. Among the class of
infinite-dimensional Lie algebras, tr1ere e;<ists a remarkable subclass_. i.e. the
affine Kac-ty1oody Lie algebras.
Introduced in 1968 by Kac [7] and t1oody [8] separate 1 y_. they
generalized Serre·'s reconstr-uction H1eorerr' (concerning U1e ser-nisirnple case),
and have been the starting point of considerab 1 y rnany studies (see ref. [ 1] and
references tr1erein for a survey). In particular. in connection w·iu-, str·ing
Hteories .. e;<plicit realizations of U1e sirnply laced affine Lie algebra~; (i.e. A~11l, Drn Efl), t t~,. 'th t"' . "b . t t· .. ' - 1 ]- . t ) ~~ ·.• ;, .!_. oge 11er \·Vl 1 11e1r as1c represen a 10n 1,::;2~ L, ror ms ... ance.
have been \·Vorked out [5], [9].. usinq as a basic tool tr1e '·/erte~: operator.
describing the emission of a "tachyon" by a bosonic string [ 1 0].
In this \·vork:_. we are interested in higher -dirnensiona 1 genera 1 izat ions
of affine Lie· a 1 gebres ; H1e aim of this paper is in fact to describe a class of
(possib 1 y infinite-dirnensiona 1) Lie a 1 gebras .. inc 1 uding as particular cases the
semisimple and the affine cases_. and to derive their general properties. We
then choose to study Lie algebras_. called the quasisirnple Lie algebras
(Q.S.L.A's) characterized by properties 'vYhicrt appear to be fairly natural and
not so rnuch restrictive :
- finite-dimensional Cartan subalgebra
- non-degenerate ad- invariant Vil1 ing f orrn
- discrete root systern
- ad-nil potency of tr1e root spaces attached to non-isotropic roots.
3
\lv'f= are U·ten able to deri'-.1e a classification theorem for the poss1ble
quasisirnp:= root systems .. and moreover to construct an e;.:plicit realization
of some CSLA's as "current algebras", generalizing the loop algebras. Trtis
approach appears to be very interesting for the study of quantum gauge
theories, 1:·1 Hte sense thEJ'c the current algebra realizations, of the form ( ·y ~ p . .T ,g,), i being the '·)-dimensional torus and Q,, a sernisimple Lie algebra,
provide us v-..-ith a nice tool to investigate the infinitesimal unitary highest
\"v'eight representations of local gauge transformations groups [ 11], [ 15].
The paper is organized as follows : in section II, we derive some
general properties of the OSLA's .. and conclude with a classification theorem
of all pos~ible irreducible "elliptic" quasisiple root systems; in section Ill,
v·le build explicit realizations of some OSLA's as current algebras ; section IV
is devotee to the conclus·ions.
4
II. DEFINITIONS AND CLASSIFICATION:
1. Definition 1:
Let g be a cornple~< Lie algebra; g is said to be quasisirnple if:
(Q.S.L.A 1) . g is pro•./ided V·iitt"t a non degenerate invariant syrnr·netric bilinear
f orrn, ca 11 ed the Killing f orrn .. and denoted by : < .. >
(Q.S.L.A.2) : g possesses a Cartan sub a 1 gebra h .. such that
_ h is diagona 1 izab 1 e
_ h is finite-dirnensional
_ Bd(h) has discrete spectrum
'with respect to ad(h), g possesses a rootspace decomposition :
h '\'':!> 9 = ® *- 9o· OCc! A .
\·vhere R = Sp [ad(h)]
and Q .. = f;-:: E g s.t. adO·t).:=<·= o~(h):=< 'r:t hE h } ·~·· .
( 1 )
t'loreover, the Killinq form induces a b"iiinear form on the dual h' of h. and the ~ .
1:-.·-·t "'""'"·u··,-.,....+ ;u- rt 1·.-. · '-'.,:, u~.o i i lp.J t..l .::. •
(Q.S.L.A.3: for any non-isotropic root C( (i.e. <ct.cc:=- ;e 0), ad(g 0) is nilpotent.
Pernar·k : it. is easil w seen u·,at sernisirnp 1 e and f<ac-tyloodw Elf fine Lie a 1 qebr·as ·- ·- ·-
are quasisirnp 1 e.
2. General P-rOP-erties :
In thi~: sub:;;ection .. \·ve enumerate a sequence of general properties the
proofs are fair! y standard and will not be mentioned here.
Theorem 1 :
i) For an~ pair· 1X.~ of roots of g, if C( + ~ ;e 0_. then -=1Jo:·· 9e.> = 0 .: on the oUter·
hand, < .. > is non-degenerate on h~<h and 9,/:Q_,:e for any 0: E R
ii) -R = R
i i i) F o r a r·: ~ p a i r c~ . ~ of mot s : [ g ... g .] c g .. ~ ·- . - 1).. (1. l_I.+J!·
If C( +~·is not a root: g .. ~= {0} - •.•.+p ..
5
iv) For any root ct. .. [9,:e9-o:] is trte subspace of h generated by ho:, hi): being the
canonical image of oc under the identification of hand h,.
~~ernark :The non-degeneraC!d of Hte f<i11mg rorrn allo\·Vs the identification of
h v~··itrt h', in the usua 1 vvay :
for any ~· E h, .. define h,, by : "
<h .. ,.. h> = X(h) ·' 'r;j hE h. ,,
Trtis allovis to carry H1e Killing form on h,:
"1 C(_,~ E h, . <C(,~> = <F1 ,hR>· - - a 1:'
Definition :2 :
A root ct. of g is called an isotropic root if <Ci.,OC> = 0
3. The non-isotroP-ic roots :
Tt-,E- non- i~;otropic roots rta· ... ·e e~.:actl y tt·te same behaviour as U"te roots
of a semis j~·nple Lie algebra (see [ 12] .. [ 13] for instance). Trten .. 1h'e vvill state,
\•vithout pn::of.. trte f o 11 0\·ving recapitulation theorem :
Theorem 3:
Let g be t quasisirnple Lie algebra, and let R be its root systern .: let o~ be a
non-isotropic root in R.
i) dim go: = 1
ii) k~ is a root"if and only if k = ± 1.
iii) for any root~ , 2 <Ci.,~>i<C(,JX> E ll
iv) for any root~~ \h,··o:(~ = ~- [2<o~ .. ~>/<c~ .. cc>] c:: is a root too.
then \·V 0:.R = R
0..1) for any root ~ .. the follovy·ing staternent is true : trtere e;<:ist tv1··o
non-negati· . .~e integral nurnbers n+. n_ sucrt Htat ~ + n c.~ E R if and only if
-n :::: n ::=: n r···1oreover, we t·tave n - n = 2 <o~_. 1_·~ :.":-/-:-..:c~,cc::-. - ·- ·. + - + 1-'
6
In the ·;ernisirnple esse .. H1ere is not ern~ i~;otropic root. and this
H·l t'- u-· r- t'- tY· 1 ,-, ·::. .-i ·=· + u- t t·l t'- ,-. ~ "j- ·=· ·=·1· f i ,-. ij- t ,. u- t"l .. , ,. I I t'- ,., f ,. r- ·=· t 1-u I i :-· ij- r-t ij- r·· rl 1 4' ] '- I . I I ' I c: ...J u ·-· \. . 1_. I ' ·J ·.J ' I '-• - ·' ·~ 11 I I I ...,. I. ~ '-· I I .
4. The isotroP-ic roots :
As pointed out first by f<sc [7] and t-1oody [8].. the infinite dirnen~;ional
,-.t ·-u·,-.f 1 "-::0 1 i.:,o:· ,·,., H(::O ·,·,:·u- tru- :·lj·,-. ,-,-,U-,f.:· . t hu.:-e- root 0:· p-o•:·•:·uC•:· •:·u- r·r··u p~rtic-·u·•] 0:.>-.,:j..,l L,.i,.:.,.:t ._ ,:r..;.~ I 1..1- ,j ~ 1_. I W I....J .' 1..11~-...J. 1..,...; ....;._sl:;;.,.J...J .J li_. '--1 I ~~
properties. \·Vr1ich ·\'ve · .. viJl point out in this subsection.
Then. we villI be in po tit ion to find the camp I ete structure of the root .s ys tern
of a quas1s1r-nple Lie allgebre, and hence to gh1e a general classification.
We begin 'Nith this firtt important lemma :
I
Lemma 4: I
Let C( be an isotropic Joot of g. Then .. for t~ny otr1er root ~ or g_.then <C(,~:..-=· = 0.
The, ,..,r-·-,u- f. ,- f t ~-,,··:· l.:.r·r··•·l-,~ i.:· b:::.·:·~·-i U- t"l t >·,..:; r·u-ll,-,.,.,,.;,.,., lilt;;. i- U I '-'' !,1 _,I~ 1q1u I..J U..;..,W '-'lr;_. I•.JH Iii~.
I -
Pro posH ion 5 :
un,jer u·:e sarrte as:;urrrption. if -=::c:z.,g,::=· ¢: 0 for a non-isotropic root~ of g_. then
~ + nIXE R .. for infinit)ely many consecutive integer n.
Proof • A·;. surne u-, at Jhere is an integer p, such that ~' = ~+piX is a root of g, i
and ~, -1x is not a root !of g. Then, let ;<~: be a non-zero elernent or g~,.
r,. "] .- g l·t-r- -nil" .- g L·"- ... ·"·1:' t::: r.· •. • o d :-~ ;·'· t::: ~:· i). . ~· ~· + '.1. • ·- • .. ( ~·
o::-et , ... - ..... ard '-' ....: [-d(" ·,f' " ...J • '\;, - ... l>'' I "'n l d . ..-'.o:' . '\l
I
Then[>=: ... >=:,] = [;< ... [>=! .. )<.]] =- ·=:)~ . ;< .. ><IX.~·'> i<. - 1), . . - 1,.1.. I u. . I..J ((. - u. . - ,_.
Cb.,nou::; 1:.; <G:~ .W> = -o~:cc~>. an,j :=< .. can be ct-uJsen in g .. in ~;uch a \·vay that ·- . • . - . - 1_1. - 1).
Tf·: en. ·.,..., .. e ~-~a..,,. e • [ :=-:: .. • :,.,
11, ] = - <c.::. . ~> ;-: ,
_,_._, . . - I
i"JC/·i. u::.i:·::; tJ1e H-iijuction proce,jure_. let u:; er:;:;urne H1at
- .-i ... ' ' ·, \' - , .• , 1 ' ... ,, (•, ·~ '··' du= ... :·=,-o) ,.·:, n-1 - -l,_t,- } ~-.C'~ .. ~:~·--· '''n-2
7
Then: [:=<_,J .. ~\) = [x_o:f· [i<.:e;<r._ 1]] I
= - [t·i_l. ;,: 1] - (n- 1 )<c~,~> i<r 1 ~- n- - ~-1
= - <fJ·i.+l'n-1 ).~ ... r.~·,_ x - (n-1 J' -=""CX. P.·";· >=' t'; ·· .v-.,.v..._,. r.-1 · ' .ol:'- ·n-1
= - n tcx.,~> xn-1
and the proposition for 0\·VS.
V·/e are no·~v in posit iorl.' to prove the 1 emma :
1 et us assume <C( .~>;:.¢0. and 1 et v =W~· .ex. · • . I · On 1? +r10:
Using U1e last propoc1
ition .. let us assume for example that the cx.-ladder of !
roots IW +ned does not have an upper bound : u·,en : n) ,_.
n~o:· 0n . -ct = ~ im v II . ·
is a lirn.it point in R .. 1
and the.re is a contradiction \NiH1 (Q.S.L.A.2) .. the root
systern oe.;ng a~sumel to be d1screte.
Tf·1en, <C(.~> = o .. and It he 1 ern rna 1s proved.
Let us denote nO\·V byi h'rP H1e real linear span of H1e roots .. and by hrP its dual
space. i
For any (;(Eh' rP·' v·..-e define 1 o: E hrP .. by :
1 (l'l.) = <CX. i'.)j !',4',1 , .. .! .. t'. .•t' I
This then induces on ~IRa symmetric bilinear form: for any ex.,~ E h.IR:
.-·1 1 .,_ - ...-c:.ll'l.-..:. (5) .... _, ((•' ~.,. - ...... ·1'='_...-
set
h* = l(h'IR)
R'. = l(R) I . .
\·\•'e can then state the follov·ling:
Proposition 6 :
< .. >· i::; n:::n-degenerate on h* ~< h*.
(6)
•
8
Proof: let 11 .. 12 be arbitrary eler·nents on h* .: then .. there e;.::ists 0~ 1 .. c:.~ 2 .:h'!P.
such 1( C:( 1 ) = 11 and 1 ( c~.,) ~ 1., .: ass urn in g that < 11 ..1..,> = 0 for· an~ 1.., in h* . ~ - ~ '
irnp1ie:; U"iat <c~ 1 .tx2:...,. = 0 .. or 1,(c:.~2 )=0 for any C(2 E h'IP .. then 11=0. ar< ~.:-,e
.... r- u- D u- ·=· i+ ,. u- r·· • ,..o 11 u-· ....... ·=· ~ • _. 1 l. i ; I I II"'""·
Clearly .. the rnapping 1. h'IP. ~ h* sends Hte isotr-opric part of Hie roots to z=:-o.
In the s.ernisirnple case, there is no isotropic root_; moreover, U1e Killing forrn
is positive definite on h'IP.·' and tt·lis al1ov·I'S the identification of Hte Cartan
suba1gebra \·vitt"l its dual.
vv·hen restr~cted to h'IP.·' is positive sernidefinite .: but let us first define:
Defln1tlon 3 :
Let g be t cJasisirnp 1 e Lie a 1 qebra .. and -=-~ .. > be its t<i11 inq f orrn : - -If -=-- -..,. ,. ·=· pu- ·=·;+ ,- •• ·e· d0 fl.t"ll.+.:. u- r·1 h' ''h' g ;.:· ·:·:=o1·u-1 tu- u._•.:. -:-.:...-,-.;.:.; •• ,-,r:' :<
-,, ~ .•, '-' . ...oil I ._, l._. (P.''• (P.·' ·l....o ._.._. ~ ...,._,l,llwlll I"''-
· • .--- ·--- ; ·=· r: u- ·= ; + ; • ' .::. ·=·.:. r·r·· ,. u-i.:. 1·1· r-,,· + e L-t r·· ·h' '' h' g ,. -::· ,-. u~ 11 .:. d c. 1 1 ; p-+ ; .-- ; ; -- ·' . .-· lw ,... ...; I 1,1 1,l ~ w.;;. I ·~;;.. L I (p'h' IR-' ~ '-' I ~ IWo \ 1 1 ~o.IL..
.•.. +L-,f! .-.+L·I.:.t- ,-.d-·:·o·:: g ;.:- ,-.d-11.:.·-11·,-.. -j:::.t·;r.,·t,:, f1) - ,,, ~o.lt.., Ul .. i a;;; L.. ...;• ... -.. lw \..o- •• ~u IIU"-J ''' -~ •
ln the fo11JV·iing. 'vV8 ··liill ahvays assurne Htat g is an e11iptic quasi:;irnpi~ Lie
a 1 qebr-a. Then .. 'l·ie can state the irnportant H"teorern.
Theorem 7 :
W1 is a finite root sy~;t_ern. called trie gradient coroot systern .. in Hie
fo110\·Ving sense:
i) : f the '•,·1leyl refl e;<ions in R'1 are defined by : T;t···..- p>=R · ....... 1 - 1 ' L,._ ·'- - • I' 1 Q:. ~' - I '•/•,•' (( .t•
tt1en R'1 i::; V·leyl- invariant.
;;· ,..,-r ·::.r·,, 1 1 .- R' 1 ·r· ./ r .J ,_,, I .:1 o: .. ·~· t:: 1·' C\-;= ·-'' ·! .--·1 1 ·--. / .--·1 1 ~-- 1· ·=· ·::. r ; n t p., r-::. 1 - • ... ! .,. IY,-·- 1 -..1 .,.1 •• ..-· ·-' I..JII !to '··-·'""i' :.,.1 1
1), t• \,I, 1.,.1, - ·-
....... , . = 1 . _ r ·? .,-·1 1 . --,. / ·==-1 1 -..,_ 1 1 '·1o.-~~. ·~, L~ .. _, ((··'~.-- • .. c< .. ':c· j ((
iii) R'1 qenerates h ... and i:; finite. ~ u·
(1): in.~ rec~?cit priv.>te disc:tJ::::::ion, '·i G. K"·~ c:.:onje.::tur~?d ttHt .3c:tu.311'J the •::h:::::: c•f qaa::i:::irnp1e Lie .~19ebn::: d indJ?finit<? ~·~PI? ·~otJld be yoid.
9
Proof :
i) is obvious
ii) 2<1 .. 1~:.::-/<1 ... 1.> = 2<et/'·>l<cccc> is an inteqral number. 0. ' J!' IJ, ' t .. ( 1:' · ·-
For any¥ E h~ IP.:
w 1 .1~(¥) = <~-(2<oc~~>i<oc,oc> )oc, ¥> a .
= [1~ -(2<C'Z,~>/ <C'(_,CX:> 1 e<) ](¥)
and trte re~:;ult follov·ts.
· ;;i) Clearifd. R'1 gener-ates h,... ~1oreover-, since the Killing form, restricted to
h,.. is positive definite, as a discrete subset of a compact set, R'1 is finite.
This concludes the proof of the theorem.
R'1 is then e finite root system .. and will allov·l the description of R.
Let J : h,..- J(h,..) = h' ... c h'm be a vector space isornorprtisrn .: \1·/e Uten define
Ute gradient root system, by (B)
.... - J(l .. , 1.-'· .. ·- 1,. 'J.
_1 • .1
vd·tere tt'ie l.i are ttre sirnple roots of R'1. On tt'te ott"ter t·t;:tnd .. if R'1 is
non-reduced.. suct1 a ct"toic:e is also valid, but one needs an a,j(litional
specifiu:t ion :
let 11, ... 10 be the simple roots of R'1, and let 10 t1e Ute unique one such
that 2 lnE R'1.
V·le set : c<j= J(lj) E R ; ttre rx.j are tJ1e simple roots cf R1 .: since R is reduced,
2rx.n is an element of R1, but does not belong toR. Consequently_. there exists
-n_ E h' "''' s J crt tJii:t t. ....--,-1 -n--,. = n 11 II\ ....._ • 1 1_1.- .... 1
fulfilling
2 O:~n + 'D E R
Tt·tjs er11j::; :.t·te ,jeterrninat.ion of J.
Let I<IR = Ker (1) ; V·/e then have :
h' [!:~ = h' * ffi kiR
and tt·tat 1J i~; trte srnall est el ernent in h' !R
( 1 0)
( 1 l)
10
Let us denote :
d. 'h' ' 1rn 1, *.J = n
d. 't. ' 1rn ~~~.IR.J = -..) then : dim (h'IR) = n + v
Every c:: E h"'IR can be decomposed as :
ex.= (oc oc ) . O' 1·
vvitt"1 oc 1E h' *·· and IX,)E kiR.
Clearly, an element of h'IR is isotropic if and only if it takes the form:
cx. = (O .. c::).
A fev·.·· e~<amQl es :
( 1 2)
( 1 3)
_ If -..;o=O R and R1 ar·e identical U"lis is the sernisirnple case .. and it is
completely solved.
_ If ··.) = 1 : this is the case· of u·1e affine Kac-f"'loody Lie algebras .: tt"1ey can be
all constructed from a generalized Cartan rnatr-i;<.
_If ..,) ~;: 2: the corresponding quasisirnple Lie algebras are nevv Lie algebras .:
they are not Kac-~1oody Lie algebras .. and they do not possess a generalized
Cartan matt-h-::_. in the sense of Kac and r··toody [71 [6].
Nmv .. using Theorem 3 .. V·/e are in position to prm1e the follo\·ving u:=:efu1
proposition :
Proposition B :
Let c:: be a non-isotropic root of Q .. and ~an isotropic element in h'IP. .: tt·ten .. if
1X+~ is a root .. so are ~ .. (;(- ~ and ~ - cc
Proof : assuming H1at cx. .. Ci+~ E R .. then C::+~+nc.~ is er r·oot if and only if
-n <. n .,-- .. , V·1ith n - n --~·-=··c/ +i', C.'-~- i .,··c,; .-,-~- - '! ---: -:: t +·' .~ - + -.._ - ... t'.• _ _.I .- .. ,,V...-· - ""-·
\·\•'e can deduce that n_ ;~:: 2 .. and trte proposition fo11ov·t;.
1 1
5. Classification of the elliP-tic guasisimgle root systems :
First, we define the irreducibility.
Deflnit ion 4 :
A root system is said to be irreducible if :
( I.R. 1) : R1 is an irreducib 1 e if :
(I.R.2): For any isotropic root 5, there exists an c<. in R1 such that c<.+o
is a root; such a root o is called an unisolated isotropic root.
Remark : It is easy to check that any elliptic quasisimple root system can be
decomposed as a disjoint union of irreducible elliptic root systems.
In the sequel, the root systems will always be assumed to be irreducible.
Let & = (oc 1 ,&0 ) be a root of g, and (IX. 1 ,c<.0 + ~0 ) be another root of g; 1 et ~o: be
the isotropic element of h"'IR defined as follows: ~a is the smallest element in
the straiqr1t line !R·!.:· such that (ct. 1 ,IX. +r ) is a root. Similarly. iii the case of a ~ YJ) I . 0 '->,1. ..
non-reduced R1 , let ~ 1 be a short root in R1, and let ~=(~ 1 .. ~0 ) be a root of g ;
we define ~-'t to be t~te smallest point in IR~0 such that (2~ 1 .. 2~0 +~-·f) E R (In .. i - -
the following, we shfll see that U"te e~dstence of ~ .. ~ is easily checked). Then ..
we can prove : I I
lemma 9:
One has tr1e three following assertions:
i) (ex. 1 ,C~0 +n~o:i) is a root for any n E 71
ii) If R1 is not-~educed: (2~ 1 , 2~0 + r~'~) is a root if and only if r is an
odd integer. /
Hi) ~~ ~ = ~~ f qr any short root ~ of R1 I
Proof: i) Let o = (cx. 1j. C<.2 +~a).
Then v•(. w (c~ 1 ,IX. +rttt r) = (c~ 1. IX. + (2+r)~ .) and <:~ a · '' 1 · o •).
(w6.wo:)m_(C<. 1,.ct0 + r~0) = (ct 1,ct0 + (2rn+r)~o:) (14)
Considering ( 14) wifh r = 0 .. 1 sho·'/vs. that (C<.0 ,C<. 1 + k~(() is a root for any
integer k. Conversel~ .. rn can be chosen in ( 14) in suet"! a way tt"tat
OC0 -~o: < tX0 + 1 (2rr"l+r)~a ~ c::.,) +~o: ,
12
which contradicts the minimality of ~-•I' if 2rn+r is different from 0 or 1
i) is then checked.
ii) Let (2~,. 2~,., + ~.' 1) be H1e root of g defined in the 1ernrna: . - r:
v\'v. (2~ 1 .2~_+L',~) = (-2~ 1 .-2 1'·.+('~) E R ~I ,. • •1,.1 ° ~I - 0 t:'1.,1 • ~I
Then (2~ 1 .. 2~,) - ~-' ~) E R and app 1 ying i) .. ii) f o 11 o··,·vs.
iii) w·2~+~·e·~ = - (~ 1 .. ~0 + ~-·~) E R
t 1 c:::: 'i • .. ...;_,
Then, from i), ~ .. ~ = n~ .. ·vvitJ1 n E IN\{o} ; R bein!~ reduced .. n i~: obviousl~ o,j,j .:
assumin!j n ;;:: 3 .. there exists n1EZ:: such tJn:!t n/4 < rn < n/2
'II {·")ll, i·' ··, .•j f•, t'4· .. ··, i - R ·,·, i~ ,.. ..•. -t-' + o:.. i~ .• = - .::. t-' + , t r 1 - n.. o:..i~. t:
i!>+m.;, ~· - . '=' - . ~
( 16)
vv·ith o < (4rn-n) < n, \·\·"t"lict1 leads to a contradiction 1N'itJ1 U1e definition of~·~.;
hence, n = 1, and iii) and the lemma follo\·V. (U1is proof is identical to
!·1c Dona li:l's proof [2]).
,jiagrarn of the gra,jient root systern t·tas to tte a connected graph.
1.·
0 ~~ = <';, = 0 cr a {V - 1 •) .-.r- <;"~ ""r d 1 ot lj'=- :=o•:··-·IJt•r.:. tt· ""-t tt· .:.r-~ '"·'·..'i·:·t ·=· .- t"l ,·.~,-,t r- r,l·,-. t_L_"_'t ~ Cl'i"·f· ••• ••• - ~~ t_l ._ •• •) 1.,.1 I r,.; ... .. ·-' ,_,._ •. ~. lr;;. I'..J. lc: .._. '=~""••·-• .. ·-' d ·=-- .. Ur -· , ··- 1• ~-.. f'J. 1 ·-' -~ ..... !
having this property. Then { 17)
Proposlt ion 1 0 :
There e>(i~;t isotropic roots o = ?· .. L. 0\EI) sucr1 that 'x~.+5 is a root of g. "U1 ~
( 1 B)
( 19)
13
that
(20)
Then one cen state:
Propos it ion 11 :
Assuming that ~cq is known, we have on 1 y two possibilities for ~o:2 :
i) ( = (o: : non-twisted case 0:2 1
iO ~ = k ~ : twisted case a2 cq
Remark: clearly, when the Dynkin diagram of R1 has only simple links (i.e. all
roots hove the same 1 ength) I there is no twisted case.
Corollary :
Let E, be on isotropic root of g_. and let e< be a root of g of minimal length .:
then e< + ~ E R
Remark : In the v-dimensiona 1 isotropic subspace of h'11:u we are not ab 1 e to
distinguist·l the different directions .: we can only precise the numtter of
twists.
The most comp 1 icated case happens when the gradient root syst ern is
non-reduced ; hence I R1 is of the form BC0 , and has the Dynkin diagram :
o----o---o- ... -o=>=e 0:1 0:~ •X! . o:,_, o:.,
where 2c.:nE R1 too. In this case, from our choice of J.. CY- 1 ..... C(0 E R, Elnd
2e<n +l) E R ; 1 et us decompose ~~ with respect to o basis of k:IP. : ,. \) i
-Q = L1 -D (21)
Obvious 1 y :
~Since R is reduced, there exists at leest one i in { 1 ..... v} such tr1at 1);;.: 0
_ Since 21X +Tl E R and 2CY.. - T! E R, tt"len V·le have ·n1 = E ~ /2 or ·o; = () n 'J n · 'J ··..:. U:n+~r
To sirnplif!-1 the notations. let us denote ~ .
14
The same proof than in prop. 10-11 leads to :
lf ~1 ,-17 l '--o:1 = LC,.
. . . ' • 1
~I - -)~I nr- 4· ~ '-.. ·:.,. rr + 11 - ~ '-.. ·- ~-..
- ··n .
denote respectively a long and a short root) : ;"'; ~i ~i ~t i .•. c.,-:.~. +' = L-, = :: ; 1. ·,en Tl = u
... •1·n ·J ' ... ~:<L ·· · ·
the root::; eJre then :
~i C\ +t1i;L
. ..,1\,· -~ . ..,n ;::i
..::..v~c- + IJ + ..::.. '-. ...,
~n E 71..
litn E 7l..
~n E 71..
·':' .. ~·!IV +l'j+ '!rlt i ,-.,· ~, / ··"'!'\' I"" ··.. - - ·") . t f·- . ..-, ~· • . •. rlt· r i .:.., 000 .:., •"•::;
0 "'-' I.,. ,' •''•L ~· ' 000 J'ol ,' •''•L J'0
- .:.., • • I I 1:! t I L 1/,::; + IXL + IJ + L I I 1:,
. .., ......... ·· ·· .. ,t_,..; r· · ·· ··· c · ..... tf· -r· ·· ·· . ...,r,..; - o
..::..·:..LIX.:.+I.:•:.l+I!+L b:, ·' .'l..-:_.>i<.IX.:·.·-~·C·;=· = L: . .II:! ''-~-L +II+ L I(,!:: r;;, ..... - ._. ·-· ~· . 1f 1=; - ·?~" 1 · ther1 -~·~ 1 = n I '"L - .:...(,. .. I I- .I .
t1i.Jt the:;e t·v·v·o :;o1utions t1ein~~ equivalent, ·vve ct·,oo:;e ·(11 = ().
Theorem 12 :
E R 'inEZ::
Let R tte en 1rreduc1tde elliptic quasisimple root system, an,j 1et R1 t1e its
~~rc11jient root :;~dstern -
i) If R1 is re,juce,j .. Hten R can be cornp 1 etel ~d cttaracterized t'Y Ute e:~:pre:;:;ion
D:::;{D •'•) T., " I, "'1 •' \.: ,1 ... ,I
·vvitJt \) = ,jirn ko;~ ar11j "'(. = nurnt,er of t'Ni:;ts
An artlitrand root takes then Ute f orrn :
~·"'r' .. fi -.. 1 tt. c 1 .·
":f'rn .. n.E2Z 1' 1
{··"')~·'! :, ..:. ·-··:
15
1i) If R1 = BCn, R is completely characterized by :
Table I :
TJ; = ~i
~ i = 2~ i C(L
i = t1+'t2+1,. .. \) E1 =4Fi , 2cx:.s+lJ ..,
i = t 1 + t:2 + t3 + 1 ... . . \.'
16
Conclusive remarks 1
: In this section, we have classified all possible
irreducib 1 e e 11 ipt ic quasisimp 1 e root systems ; however, this c 1 assification
does not e:<tend to a camp 1 ete c 1 assificat ion of the quasisirnp 1 e Lie algebras.
\11/e have not proved that the root system completely determines a Lie algebra;
this is probably related to the fact that in the case\)~ 2 (tr1e case 'J = 0 or 1
are well known), the 'w'ey1 group is no longer a Coxeter group (see [ 13]).
Nevertheless, it is possible to compute the root-multiplicities, and to prove
that an irreducib 1 e ~ 11 ipt ic quasisirnp 1 e Lie a 1 gebra of type \' possesses a I .
\,;o-dirnensiona 1 centler ; it is possib 1 e to investigate the possib 1 e Killing farms.:
for such an ana1ysif, we refer to [ 15].
I I
Ill. CONSTRUCTION : '
After having classified them .. ·vve are able to prove that some of the
new infinite-dimensiona 1 Lie a 1 gebras exist. '•fle can build irreducib 1 e e 11 ipt ic
QSLA·'s for arbitra y \.' and t = 0, 1 or 2 .. using techniques familiar from the
affine case, i.e. realisations as current algebras (see [ 1] for instance).
1. The non-twisted case :
Let g0 be a iimple complex Lie algebra, with Cartan subalgebra h0 , and
root system R0 ; 1 ~t { oc 1, ... e<.0 } be the simp 1 e roots of R,) .: 1 et T\) be the
\)-dimensional torts, provided with the Lebesgue measure dt, such that:
J T'Y dt = 1 I - (24) i
Let g 1 = P(T\),g0 ) b~ the Lie algebra of functions wlth finlte Fourier series, I
with the point wis~ Lie structure : 'V x,y E g 1
[x,y]1 (t) = ~x(t),y(t)~ '7 t E T\)
i <.-x:,y> 1 = II '> <:><(t),y(t)> 0 dt (25)
. 1f I
{2): Actually .. V.G. K.ac syggested th.:1~ for '>>2, there could be m.aniJ Lie algebras corre:spondk19 to) a 9ivo?n root system ; this is .an i1teresting still open problem.
17
In order to get a finite-dimensional Cartan subalgebra, we define the
derivations D(d) of g 1 (d E [\)) by : 'r:t )< E g 1
D(d).x = (d.~).x (26)
and g2 to be the semidirect product of g 1 by [.\) ; I
witt1 commutation relations : !
[(x_.d),(x',d')~ = {[~<,x'] 1 + D(d).~<' - D(d').x , 0)
A Cart an sub a 1 gebra of g2 can be written in the f orrn : 1\)
h2 = h0 III L,~ [d;
(27)
(28)
where the d; 's forr-rl a basis of 71..\).
Finally, to g~t a Killing form, we define on g1 the following family of I
[-Ve I ued ~il ine~r f,trms .:
-.v/x,y) - <x~D(d).y>1 i
Propos it ion 13 : I
'r;/ d E [\)
":/X 11 EQ ~~ 1
For any din [•, 'I'd it a Chevalley 2-cocycle:
~: -.v,/x,y) =-r,/y .. x) ":/ X_,y E g 1 , .
11) l¥"i (x,y],~) + V/(y,z)lx) + 'lfd~ [z .. xl,y) = 0 ":/ X_,y,z E g1.
(29)
The proof f o II ows Jasil y from the inveriance of < , > 0 . Consequent I y, it is
possible to make a ~entral extension with respect to the -.vd ; let then C be a I
v-dimensional cen~er',. spanned by complex linear combinations of the vectors
c1, ... c\) . Then, we ~efme a L1e algebra: I
g = P(T\) ,g0 )1 ® L: [d; (£! L.; [c; (30)
w·ith Lie bracket: / ":/ {x,d,c},{~<',d',c'} E g
[ { x ,d,c}, { x',df ,c'}] = {[x,x' ]1 +D(d).x' - D(d') .x , 0 , L.; l¥'d/x,~<.').c,} (3 1 )
It remains to deflnt- the Killing form; we set: 'r;J {x,,j .. c}.. {x',d',c'} E g
<{x,d .. cL{>f',~',c'}> = <x,x-'> 1 -d.c'- ,j'_c
where . is the usua 1 seal ar product in f~.
18
Proposition 14 :
< .. > is a non degenerate syr-nrnetric invariant bilinear form on g.
Proof. non degeneracy and syrnrnetry are obvious:
L.:. t r .. ,.. ..; .- 1. r ...... • d·' -.'1. r ..... " d" ,-. "1. .: g "'~ l., .. ,u ....... J.• i.'··.. ,L ·'' i.."· .• , .... ·' _
[ r ,, d -. 1 r. ,.. d' ·'1 ] f,." d.. -." 1 ...... < 1.··\, .•{.: .i 1 '!,i<, .• ,c f .' 1."• I J {.: f /
- -~ ["' ,_ ... , [itd'l ,,.. Ol'd·'i ') 1) ,.~ ·,1( i'v ····'i -. 1. rv" d" c"1. ~--- ··- .•. ·' ,., J 1 + '· .• . ,.-., - ' ' . ,., ' ' .;_, 1 T ,j · ._,., ',., ... {.: ;J ' i.. ,., ·' 1 .I ,. 1
- ·· [•.t • ·'] •.;"~ { v" ,,·,, ·1, { ,." 'J ·,} lr ( 'ol X'.'J - -= •. ···.··" 1' , , .... 1 + l~f d ,,, .. X ·' - 'tid'·.<'· , ..... - lf d"'' ., ..
,·· f ' ' j -. l [ f.¥... j' ·' 't f •,/' j" . "l] .. ··-1. ·'"·.·' ,'-·.r .. l''' .•' _.C .1 _. ... ··' ,C .I ~ ..
{ j l f [ ' "] c· 'j··· .. D .. d ... , · ·- ~···;· ,. · .... ·t = < ::<,1 .•CJ , I. X .• X l + .11,1 . ,t.X - ~~ }.';{ _. 1.) .• '- 1 l~id/'K .. :..: }.C11>
and tr1e propqsition follo'l·vs from trte invariance of<. >1.
<, > i~:: tt"1en a rele'.. .. i:Jt"it Killing forrn for g.
J't"1e Cart an sub a 1 qet1ra takes tr1e f orrn : I ~
- ,. ~· . - "'\' \) -. h - h0 (fl ,;;;_ 1 [d; 1.±1 ,;;;_ 1 [L;
i:lnd \·ve 8i:lSi1 !d see tt·1at g is a DS.L.A ..
~~ · - r ,.. .. · •· .- " - t - .-. t I' j 1• 1· t ,... r-- - t .-·11·-· t -rr 1· - t t· - .-. r - -. t r-l' r· · -f "" 1 { h 1 ,\'I:! 1•.1 ';";' ; !ij '•' I:! .u ·=· .. ~I .':j • .:· uu. ·=·;,·=· '= I, .I:! .• It=. ·=·,..!t=.l_. ~ fl u I.JIJ '· ·'
\1 c E c. ad(c) = o L -• 11 - "'\' ') 1 j . . ..... - h ·:. • .. • - .- 1 ( 11 - [1 ( j ) - '='· .• - "-1'\; I i ' .~1:!. I.J,I:! •jL .. L .. - ... 1 ·'
and V·ie ha'·/8 to soJ•.,..e :
D(d).X = rn(d) X Vv'itt"t X E g
Tt·1e solutions are of tt·1e form :
l . . [n:"' . ,,E Xm : t E T ~ Xm{t) = _1 e:=~P{irnkt~)J·
··r1 r ri'' - 1· "'\' ··;· ·~. r~l ;, •.'J, - ... 1 '''k l'k
· .. v h P r-P ··q = r rn r1 ' .: 71·,;· I - ._ ·' 1. 1' ... '1~'1·' ._ U;.
foran11hEh: .::\ I)
' • • I .. ·- dP't' v 1 - · .. · t t· I v •j •. t.,.. ..... m,o:- 1. ..... 1 .• l'·m_..:(
'Mt··- ,-- }I - [ rrv -• .. •r 1it•r t ''·] E . E .: t'g ., .''' ,I:!, 1:' ···rn,o: - , · 1 t',····l-1 \. • 1k "k·1 • o: ' o: ·- '· o' o:
{'"J:;'"J:; 'j '· ·-· ·-· ..
(34)
(35)
19
The spectrurn of adO""!) is then given by : 'r:/ {h,d,c}.: h
d' (L 1"1 1' " [ ',-!' 't '] '' a r. .. ,,, • u • c i ·' J:... .. = m r.. u ,t + 1X '·· l! ,':·r.-. .. • ' ' · !II,: IJ. 11_, I,J.
(36)
The Killing f orrn is canonica 11 y carried onto h. by defining, for any ¥ E h, .. the
canonic a 1 identification.
-:·· u ( .,. 'I L-~··:- - .,. { L·l 't '"' •. ~ , , I , - ~ •. I ,
Then .. if Cl + rn is any root of g : - { r.r r .... ,, - r ... (l 1. '\ . ..,. r· -. 1. t:! ·.'-'~ + II_. - i. ''o: .. - . ..:...1 rlkL:k J
(37)
(38)
and rn is easily identified es being the isotropic pert of the root c<. + m .
t·'1oreover. it is eesy to check U"1t1t < .. > is positi'-.1e semidefinite of corank \) on
h,!P.i<:h,!P. .. and it follows:
Theorem 15 :
g 1s an elliptic quasisirnple Lie algebr-a .. associated '·/·lith the root systern: •
R = (R_ ; \). 0 ) IJ ' •
Any non-t·.,...,.·jsted e11iptic C! S.L.A. can then be r-ealized in t~1is \·vay.
2. The case 1: = 1 :
The construction given tr1ere is fairly standanj, and can be found for instance
in Kac [ 1].
\11/e consi,jer a simple Lie al,~ebra g0 _. \ollt"IOSe Dynkin diagram admits a non
trivial automorphism 1l_. of or-der k .: 1 et now Q be an automorphism of order k
of the torus T'"' .. for e::<:arnple :
t, ~ t, + 211/k
'r;j i ;C 1
'v·le define the autornorphisr·n c/ on g by : ...... /'
, f I I '\o • (\I 'II
a ' .. Y..J = a '·.1·. o P.-·'
c/(<"·)=c . '--; .• .. i '·.·· a(d.',,=d.
. 1 1
w· .. ·~g '•' i\ t 1
'r;j i = 1 ..... \)
'r;j i = 1, ... ··.)
(39)
20
.., ,...,. Let g be the fixed point set of cr ; we have :
.., "' k P 'T . .,.. ( -4 J' "' ..,.. [ d ' ..,.. [ 9 = "-'1 i1' .·9.:. ..' !±l ..:;..., ; !±l ~, c; (40)
where g~J = { x E g0 / Q.X = exp (i2n:l /k) x } -c:.nd P (T..,.. (1)) 1 ~ P'T..... (-l)) t [ '•·)] (t) u 1 ,Q 0 = \ :><: t ~ , Q0 S. . cf~r\. = X
There are five possible automorphisms for the g0 given by the five possible
QSLA's (R1 ; '), 1) or (BC0 ; \); 1, 1 ; 0, 0) .: in each case, it is easily seen
that g~ol is a simple Lie algbra:
A ·2 1 k·l B - 9o = 2n : a = g 1) = r.
R(ij):::: (BC0 ; \); 1, 1, 0,0) 2 (o)
- 9o = A2n-1 : ct = l 9o = Cn ..,
R(g) :::: (Cn ; v ; 1) .., fo l
- 9.:. = Dn+ 1 : cr"' = 1 9~~ . = Bn ..,
R(g) :::: (BC0 ; 'J ; 1, 1, 0 ,0) 3 (o l
- gl) = 04 : ct = 1 9.:.. = Gz ,."".. f ) R~9 J :::: ~G · .., ) ·. 1
0 2 } 'rl' ·' }
- g = E : cr2 = 1 9lol = F Q 6 1) 4
R(g) :::: (F 4 _; \} .. 0)
(The proof of this result can be found in Kac [ 1] )
This leads to the follO\·ving:
Theorem 16 :
Any irreducible ellipttc quasisimple root system of twist on can be realized
as a "generalized loop algebra", as constructed in section 111-2 .
..., 3. The case 1: = 2 (reduced R1)_.;.
In this subsection, we shall see that it is possible to build a Q.S.L.A. for any
irreducib 1 e e 11 ipt ic quasisirnp 1 e root system .. very sirnil ar to the one that
1 eeds to the t v·..-ist one Q.S.L.A's.
21
First of all, let 9,) be an affine Kac-Moody Lie algebra(\) = 1 ), and let
us define : ,.. 9,:. = 9,/[d l@[C 1 (41)
with the usua 1 no tat ions.
It is fairly easy to see that
~~ ..... ) ' \) . ' \) ( ) g = P(T ,g0 ® "'-l [dl ® "'-l [c1 A2
provided vvith the usual Lie product and Killmg form is an elliptic Q.S.L.A. ,
with root system:
R :::: (R1(g0 ) ,; \), 't(g0 )) ('t = 0 or 1)
if R1 (g,) is reduced, and the corresponding one if R1 is non-reduced :
R :::: (BCn ,; \) ; 1, \)- 1, \)-1, 0).
Now, 1 et cr be the straigrtt e~<tension [ 12]of an automorphism of the Dynkin
diagram of g0 .: it is easy to see that cr acts trivia 11 y on c 1 ~nd d 1 ; 1 et k be the
order of cr, and let Q be an automorphism of order k of T"'- 1.
'v·le define the automorphism fi' of g by :
~ .... ·(v' ·(v ) a ·''· ..' = a .. 1\ o Q,
(i'(c.) = c. 1 1
'r:/ X E P (T"'- 1 , g)
"i/ i = 1 ,. .. \)-1 ,.._.. a·td.) =d. 'r:/ i = 1 .... \)-1
\ 1' 1 ,
~ 'v and g is the fixed point set of cr; it is a Q.S.L.A., and we will look for its
root system.
1 n the seque 1, we ·will use the notatlons :
E1, Fi' ~~are the Chevalley generators of Q0
~1 are the roots of g. . . ...,
e .. f .. 0~.yare the Chevalle" generators of g 1' 1' 1 :j .
~
oc1 are the roots of g.
(43)
i) 9,) = (Dn+2; 1, 0):
cr(~ 1) = ~o ; a·c~) = ~ 1
cr(~n+1) = ~n+2 ;cr(~n+2) = ~n+1
V•le set :
.... v C(i = ~i+ 1
.. ,. .. , v ocn = ~n+ 1 + ~n+2
22
~ /~t\·~
r,··~~ lo ~~••
· 'P, ) f~ (" ·"' ' a l.t:'1 = t'; .. l=L, ... n}
e = E + E1 0 0
8; = Ei+ 1
f =F +F 1 0 0
'. -') ) ',1=..:.. ..... n
8n = En+1 + En+2 f - F + F n- n+1 n+2
These ore the Cheves 11 ey generotors of
g{o) = (B · 1 1) o n ' ·'
.., ~ow .. using the ono1ogue of (40)_, we are ob1e to build the corresponding g_. end
its root s~stern rev eo 1 s to be : ...., .
R(g) = (80 _; -v .. 2)
l.l.i g - I'D · 1 fl) ' c' - •. 2n } } ....
. ' 2 1 = u, .... · n
\¥e set : v v v
~; = ~i + ~2r.-i e.= E.+ E ..... 1 1 Lfl-1
These ore the Chevalley generotors of
g~o) = (Cn _; 1, J) ..,
It is then eosy to find the root system of g
iii) 9.:. = (EE. _; 1 I 0)
cr<~2) = ~2 cr(~.) = ~~t- cr(~E) = G•4 _; (J"(~4) = ~o
cr(~ 1) = ~5 ; (r(~5) = ~3 _; cr(~3) = ~ 1
f.= F.+ F2 . 1 1 n-1
I
f-
23
.. ,. ~: ~v v e = E + E4 + EF Ct. = + + ~6 1) -4 0 1) -·
..,. ~·; +
·y· 'of
e 1 = E 1 + E3 + E5 (:( = ~3 + ~5 .., .....
" ~v e..., = E..., Ci....., = 2 ~ i.. .:.
are the Ct"tevall ey generators of
g(o) = fG · 1 1) 0 ' 2 I I
..., and we easily check that R(g) ~ (62 ; 'vI 2)
iv) g0 = (E7 ; 11 0)
cr(~;) = ~6-1 (i = o, ... 6)
cr(~7) = ~7 Then
e.= E.+ E6 . 1 1 -1
'y' • v c::4 = ~7 84 = E7
are the Crtevall ey generators of :
g(_o) = • , ~-~ ~F 4 ; 11 1 J
and vv·e get :
-R(g) ~ (F 4 ; \)I 2)
i) - iv) lead to the following:
Theorem 17:
fl) = F 0 + F 4 + F 6
f 1 =F 1 +F3 +F5
f...,= F..., ~ L
f1 = F1 + F 6_; (i=0,. .. 3)
f4 = F7
Any irreducible elliptic quasisimple root system, with reduced gradient root
system, of twist two, can be realized as twisted extension of an affine
Lie algebra.
IV
V·/e will now· get a sirnil ar result in the case when R1 is not reduced.
24
,., 4. The case R1 non-reduced :
i) First of a 11 ~ take Q0 = (BC0 ; 1 ~ 1 ~ 0,0 ,0) ~ and make the non-twisted ,..,
extension g of (42) ; the root system of g is clearly : ,..,
R{g) = i'BC · 'J · 1 \)-1 \.)-1 0) ' .. n .. " ' .• }
The other QSLA's ¥tith non-reduced gradient root system will be built making
twisted extensions of affine Ue algebras~ llke in the subsection 3.
ii) 9o = (A2n+1 ; 1' O)
0'(~,)= ~2n+ 1-i
~: = 2(~; + ~;n+ 1) eo = 2(Eo +E2n+ 1)
or:~ = 2(~~ + ~~+1) en= 2(En+En+1)
oc~ = 2(~: + ~;n+ 1-i) e; = E;+E2n+ 1-i
are the Cheva 11 ey generators of :
g(o) = {6 · 1 1) o ' n ' ·'
f = F + F o o 2n+1
f1 = F; + F Zn+ 1_1 (i=2 .... n-1)
g~1) is an irreducible g~ol_module~ and the roots of g are .. by a simple
Then:
iii) 9o = (C2n+ 1 -; 1 ' 0)
cr(~i) = ~2n+ 1-i
'r;/ n1 E 71
'r;/ n. E 71 1
'r;/ n. E 71 1
· f = F +F n n n+ 1
are the Chevalley generators of the affine Lie algebra
g~~:)) = (BC0 .: 1 ; 1 .. 0, 0~ 0)
25
The usua 1 ca 1 cu1etions 1 ead to : ...,
R (g~:::: fBC ·9 · 2 ,.,_.., .... v-'? 0) . ~ ' n .} ~ } , } -.·
"' (i=O, ... n) ' (h---{)---0-- . . . ---{}:;;=> ::aO ~
-~~+1 ~,., ~~-
e.= E.+ E..... . f.= F.+ F2 . (i=O, ... n-1) 1 1 .!.1"1-1 1 1 n-1
v v .. , OC; = ~; + ~2n-i
e = E n n f = F n n
are the Chevalley generators of : ~1 ( ) 90 · = .BC0 ; 1 ; 1, 0, 0, 0
Tr1e exp 1 ic:it cal culet ion of the root system 1 eads to : ...,
R(g) = (~Cn; '); 1, ')-1, v-2, 0)
,.. - 2 ) '..1 = j, .... n.
v v v v y CY..o = ~o + ~1 + ~2n+1 + ~2n+2 eo= Eo+ E1 + E2n+1 + E2n+2
f = F + F 1 + F2 1 + F2 2 o o n+ n+ y v v
CY..rr = ~i+ 1 + ~2rr+ 1-i e. = E. + E2 1 . f. = F 1 + F2 1 . 0=2, ... n- 1 )
1 1 n+ -1 1 rr+ -1
v y CY..n = ~n+1
are the Chevall ey generators of :
(o) - fBC · 1 1 o 0 0) go - •. n } 1 •• "'··"" 1
The extension leads to : ...,
R(g) = (BCr. ; v ; 1, \)-2, v-2, 0)
26
The result of this subsection is an e~<tension of Theorem 17 to non-reduced
gradient root systems :
we get realizations of : (BCn; '·) ; 1, \..:0-1, v-1, 0)
(BCn ; v ; 1, v-2, 'J-2, 1)
(BC · v · -? \.)--? 'v-2 0) n J } ~, *J '
(BCn; 'J .: 1, v-1, 'J-2, 0)
(BCn ; \) ; 1, \)-2, \)-2_. 0)
In particular, specializing our results to the case 'J = 2, and referring to the
classification Theorem 12, we get this rtice conclusion:
Coro11ory :
All the irreducible elliptic quasisimple root systems of type 'V = 2 possess a
realization as "genralized loop algebras", or "current algebras" ... as described
in sect ion Ill.
IV. CONCLUSIONS :
The results of section Ill show that at least some of the quasisirnple
root systems classified in section II possess a realization as as current
algebra; unfortunately, due to the lack of a generalized Cartan matrix when
\) ~ 2, we have not been ab 1 e to genera 1 ize such a construct ion to twist three
QSLA's fcir instance. Nevertheless, we believe that it is possible to associate
a quasisimp 1 e Lie a 1 gebra to anyone of the root systems c 1 assified in
Theorem 12 (and perhaps many QSLA·'s for a given quasisimple root system
for -..) > 2, as suggested by \I.G. Kac) ; ab8tract constructions are currently
under study.
27
Another very interesting application of the theory of QSLA .. s is the
representation theory, which is probab 1 y c 1 ose 1 y connected to tr1e
representation theory of the gauge groups, in quantum gauge theories.
Actually, some unitary representations of the gauge group, called the energy
representations, have been accurately studied [ 16]. [ 171 but very little was
known about highest weight representations. The quasisimple theory allows
the study of such representations [ 11]. [ 15] ; in particular they exhibit, as a
representative of the central extension terms (see sect ion Ill), an expression
identical to the Schwinger anomalous term in quantum current algebras [ 18].
This then appears to be a very promizing field of research, in connection with
quantum gauge fie 1 d theories.
Aclcnowl edgments :
It is a pleasure to thank Pr. S. Albeverio, Pr. V.G. Kac, Pr. M. Sirugue
and Pr. D. Testard for fruitful discussions.
L
28
REFERENCES
[ 1] V.G. Kac : ·Infinite-dimensional Lie algebras", Cambridge University
press ( 1985).
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[ 1 0]
[ 1 1 ]
[ 12]
[ 13]
[ 14]
[ 15]
[ 16]
[ 17]
[ 18]
I.G. Me Donald: Invent. Math. ~ ( 1972) p. 91
V.G. Kac: Adv. in Math. 30 ( 1978) p. 85
Date, Jimbo, ~ashiwara, Miwa: J. Phys. Soc. Jap. 50 ( 1981) p. 3806.
I. Frenkel ; V.:G. Kac : Invent. Math. 62 ( 1980) p.23
J. Mickelsso~: Comm. Math. Phys. 97 ( 1985) p. 361
V.G. Kac: Ma~h. USSR 1 zv. 2 ( 1968) p. 1271 i
R. V. Moody: 1· Alg . .1Q ( 1968) p.211
P. Goddard, D~ Ollve: MSRI publlcationsJ., Springer Verlag ( 1984) p.51
J.H. Schwarz : Phys. Rep. 89 ( 1982) p. 223
B. Torresani :1 preprint CPT 86/ P. 1904, to appear · ·
in Lett. in Math. Phys.
H. Freudenthrll, H. De Vries: "Linear Lie groups". Academic
Press ( 1969
N. Bourbaki : "Groupes et algebres de Lie", chap. 4, 5, 6.
Elements de mathematiques, Herman I
E. Cartan: T~esis ( 1894)
B. Torresami : "Representations projectives des groupes de
transformat~ons de jauge locales", dissertation,
preprint CP~ 86/p. 1893
I.M. Gel'fandf M.l. Graev, AM. Vershik: composition math. 35,
3, ( 1977), P-i 299
S. Albeveriol, R. Hoegh-Krohn, D. Testard: J. of functional analysis~ I
( 1 981 ), p. 3f8.
I. Bars: MSRI publlcations,l Springer Verlag ( 1984), p. 373.